CN105657216B - Component for low smooth noise reduction filters - Google Patents
Component for low smooth noise reduction filters Download PDFInfo
- Publication number
- CN105657216B CN105657216B CN201610124067.5A CN201610124067A CN105657216B CN 105657216 B CN105657216 B CN 105657216B CN 201610124067 A CN201610124067 A CN 201610124067A CN 105657216 B CN105657216 B CN 105657216B
- Authority
- CN
- China
- Prior art keywords
- pixel
- filter
- sef
- image
- region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000009467 reduction Effects 0.000 title abstract description 5
- 238000001914 filtration Methods 0.000 claims abstract description 29
- 238000000034 method Methods 0.000 claims description 27
- 230000000877 morphologic effect Effects 0.000 claims description 7
- 230000000052 comparative effect Effects 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 2
- 239000004744 fabric Substances 0.000 claims 1
- 230000003287 optical effect Effects 0.000 abstract description 13
- 238000010586 diagram Methods 0.000 description 13
- 238000004458 analytical method Methods 0.000 description 11
- 238000003708 edge detection Methods 0.000 description 6
- 238000005215 recombination Methods 0.000 description 4
- 230000006798 recombination Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000002845 discoloration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
- H04N5/21—Circuitry for suppressing or minimising disturbance, e.g. moiré or halo
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
- H04N5/142—Edging; Contouring
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/10—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from different wavelengths
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/80—Camera processing pipelines; Components thereof
- H04N23/81—Camera processing pipelines; Components thereof for suppressing or minimising disturbance in the image signal generation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20004—Adaptive image processing
- G06T2207/20012—Locally adaptive
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Color Television Image Signal Generators (AREA)
Abstract
The present invention relates to a kind of component filtering for low smooth noise reduction.Generally speaking, in one embodiment, it is respectively filtered by luminance component to image and chromatic component, low optical noise is removed from image by being based at least partially on the Gaussian Profile of image adaptively to image filtering, and/or by dividing an image into separated region and respectively filtering to each region.
Description
It is on November 15th, 2011 that the application, which is application No. is 201180062954.X, the applying date, entitled " for low light
The divisional application of the application for a patent for invention of the component filtering of noise reduction ".
Related application
The U.S. Patent Application Serial Number 12/950,664 submitted this application claims on November 19th, 2010；In November, 2010
The U.S. Patent Application Serial Number 12/950,666 submitted for 19th；And the U.S. Patent application sequence that on November 19th, 2010 submits
The priority of row number 12/950,671, the complete disclosure of these patent applications is incorporated herein by reference.
Technical field
The embodiment of the present invention relates in general to video frequency signal processing, and more specifically to processing video letter
Number to remove the illusion as caused by low optical noise.
Background technique
Low light image is particularly susceptible the noise caused by light detecting sensors (that is, low smooth illusion) due to deteriorates.For example, video
Or camera may capture unexpected particle or discoloration in low smooth situation.This noise may cause incoherent pixel,
And thus leading to the compression efficiency of video coding algorithm (for example, MPEG4 and H.264) reduces.(such as safety is taken the photograph for many applications
As device) it captures low light image and needs the memory space of enormous amount to be used to store these images, and required memory
Any reduction spatially all may cause more cost-efficient application, and the image of storage or the quantity of video frame increase or uses
It is reduced in the Internet traffic for transmitting these images.Therefore, there are many work to be dedicated to detecting and eliminating low optical noise.
But there are scarce for previous work (such as transform domain method, DCT, small echo (wavelet) or other statistical methods)
Point.These methods are to calculate upper intensive and need significant a large amount of computing resource, in this way for low-power portable or other dresses
Setting may be unavailable.It, further will money furthermore these methods can not be adjusted based on the complexity of available resources or source images
Source is wasted on simple image or during it may not be needed additional resource or do not have the available high-load condition of additional resource
Waste of resource.
Summary of the invention
Generally speaking, the various aspects of system and method described herein are using Gaussian Profile and correlation technique come from taking
From removing incoherent low optical noise in the image of video or camera.These images can be separated into brightness and coloration point
Amount, and respectively it is filtered.According to the complexity of image and available resource, different filters can be used.This
A little filters can adapt to the variation in image by using edge detection and expansion filter, thus keeping characteristics edge
High frequency detail.Furthermore multiple portions can be divided the image into, it is respectively filtered and is recombinated.
Generally speaking, in an aspect, a kind of system for removing noise from low light image includes luminance filter
Circuit, chrominance filter circuit and adder circuit.Luminance filter circuit is by applying first to the luminance component of low light image
Filter creates filtered luminance component, and chrominance filter circuit by chromatic component to low light image using the
Two filters create filtered chromatic component.Adder circuit is by filtered luminance component and filtered coloration
Component combination, to generate filtered low light image.
In various embodiments, luminance filter circuit selects first filter from multiple available filters, and should
Selection can be based at least partially on the calculating cost of each filter in multiple available filters and/or the complexity of luminance component
Degree.First filter can be low pass mean filter, median filter and/or the sef-adapting filter based on edge detection；
Second filter can be low pass mean filter and/or time average filter.Image can be still image and/or video
Frame.
Generally speaking, in another aspect, a kind of method for removing noise from low light image includes by low
The luminance component application first filter of light image creates filtered luminance component.Pass through the coloration point to low light image
Amount creates filtered chromatic component using second filter.By filtered luminance component and filtered coloration
Component combination, to generate filtered low light image.
In various embodiments, first filter is selected from multiple available filters.The selection of first filter can at least portion
Divide ground based on the calculating cost of each filter in multiple available filters and/or the complexity of luminance component.First filter
It can be low pass mean filter, median filter and/or the sef-adapting filter based on edge detection；Second filter can be with
It is low pass mean filter and/or time average filter.Image can be still image and/or video frame.
Generally speaking, in another aspect, a kind of sef-adapting filter for filtering out noise from low light image includes
Morphological filter and compare filter.Morphological filter is divided an image into non-by the edge swell that will be detected in image
The fringe region of fringe region and expansion.Compare filter for the region ratio around the pixel and the pixel in non-edge
Compared with, and the pixel is optionally replaced with into the new picture at least partly obtained from the region around the pixel based on comparative result
Element.
In various embodiments, the edge in difference filter detection image and/or morphological filter include expansion filtering
Device.Expansion filter can be 3 × 4 expansion filters, and the region around the pixel can correspond to 3 × 3 pixel regions.
Sef-adapting filter can be by the region around poor and pixel between the average value in the region around pixel and the pixel
Variance compares.The variance can be the variance of the Gaussian Profile in the region around the pixel；It may include a kind of circuit to calculate this
The variance of the Gaussian Profile in the region around pixel.The average value being averaged for the Gaussian Profile in the region around the pixel
Value；It may include a kind of circuit to calculate the average value of the Gaussian Profile in the region around the pixel.
Compare the area around the intermediate value and/or the pixel in the region that filter can be based at least partially on around the pixel
The low-pass filter values in domain obtain new pixel.New picture can be replaced with for the second pixel of the neighbouring pixel by comparing filter
Element；The pixel can be even pixel, and second pixel can be odd pixel.Alternatively, comparing filter can retain
The pixel.
Generally speaking, in another aspect, a kind of for adaptively including from the method that low light image filters out noise
The edge that detects in the picture is expanded to divide an image into the fringe region of non-edge and expansion.By non-edge
In pixel compared with the region around the pixel.Based on comparative result, optionally by the pixel replace at least partly from
The new pixel that region around the pixel obtains.
In various embodiments, edge can be detected in the picture.It can be with compared with the region around the pixel by pixel
Including by the mean of variance ratio of the difference between the variance in the region around the pixel and the pixel and the region around the pixel
Compared with.The variance be can be the variance of the Gaussian Profile in the region around the pixel, and and the average value be can be the pixel
The average value of the Gaussian Profile in the region of surrounding.The intermediate value in the region around the pixel can be based at least partially on and/or be somebody's turn to do
The low-pass filter values in the region around pixel obtain new pixel to lead.Second pixel of the neighbouring pixel can be replaced with newly
Pixel,；The pixel can be even pixel, and and second pixel can be odd pixel.Alternately alternatively, comparing
Filter can retain the pixel.
Generally speaking, in an aspect, a kind of system for removing noise from low light image includes dividing circuit, filter
Wave device circuit and recombination circuit.It divides circuit and divides an image into multiple images region.Filter circuit passes through to multiple images
The luminance component application first filter of each image-region in region creates multiple filtered image-regions.Recombination
Multiple filtered image-regions are combined into filtered image by circuit.
In various embodiments, filter circuit is once to an image-region application first filter.Alternatively, filter
Circuit can be once to more than one image-region application first filter.Image-region may include square piece, rectangle
Piece, row or column.First filter can be low pass mean filter, median filter and/or sef-adapting filter；Adaptively
Filter may include morphological filter and/or compare filter.Second filter can be by each of multiple images region
The chromatic component of image-region filters, and recombinating circuit can be by the filtered luminance component of each image-region and each
The corresponding filtered chromatic component combination of image-region.Recombination circuit can store and image block, image line and/or figure
As arranging relevant historical information.
Generally speaking, in another aspect, a kind of method removes noise from low light image.Divide an image into multiple figures
As region.It is multiple through filtering to the first filter creation of the luminance component application of each image-region in multiple images region
The image-region of wave.Multiple filtered image-regions are combined into filtered image.
In various embodiments, sequentially to each image-region application first filter.Alternatively, can be concurrently to multiple
Image-region application first filter.Using first filter may include image-region is filtered, to image-region into
Row median filtering and/or image-region is adaptively filtered (its may include control adjacent pixel carry out compared pixels, and
And optionally replace the pixel).It can be to the chromatic component filtering of each image-region in multiple images region.It can will be each
The filtered luminance component of image-region filtered chromatic component corresponding with each image-region combines.It can be with
Store historical information relevant to image block, image line and/or image column.
By reference to being described below, drawings and claims, these and other targets of present invention disclosed herein connect
It will be apparent from advantages and features.Further, it is understood that the feature of various embodiments described herein and non-exclusive, and
And can exist in the form of multiple combinations and arrangement.
Detailed description of the invention
In the drawings, similar quotation mark generally refers to identical part in multiple and different views.It retouches below
In stating, various embodiments of the invention are described with reference to following accompanying drawings, in which:
Fig. 1 is according to an embodiment of the present invention for removing the block diagram of the system of noise from low light image；
Fig. 2 is that diagram is according to an embodiment of the present invention for removing the flow chart of the method for noise from low light image；
Fig. 3 is the block diagram of sef-adapting filter according to an embodiment of the present invention；
Fig. 4 is the example of low light image component according to an embodiment of the present invention；
Fig. 5 is that diagram is according to an embodiment of the present invention for adaptively filtering out the process of the method for noise from low light image
Figure；
Fig. 6 is according to an embodiment of the present invention for dividing image with from the block diagram for the system for wherein removing low optical noise；
And
Fig. 7 is that diagram is according to an embodiment of the present invention for dividing image so as to from the method for wherein removing low optical noise
Flow chart.
Specific embodiment
Fig. 1 diagram is for removing the system 100 of noise from low light image.As the skilled person will appreciate, may be used
Source images 102 are separated into brightness component 104 and color component 106.Brightness component 104 is alternatively referred to as Y or brightness point
Amount；Color component 106 is referred to as UV or chromatic component.In one embodiment, respectively right using different filters
Brightness component 104 and color component 106 filter.Once brightness component 104 and color component 106 are filtered, then
It can combine them to re-create the filtered version of original image 102 or be further processed as separated component.
The network of switch 108 selects brightness point of one of three filters 110,112,114 for image 102
Amount 104.But system 100 may include any amount of brightness component filters, including a filter, and this hair
The bright filter for being not limited to any particular number or type.In one embodiment, if if source images 102 are simple, only need
The filtering of small degree and/or if system resource is limited, switch 108 can choose low pass mean filter 110.Low pass mean value
Filter 110 makes the high frequency signal attenuation in brightness component 104, while low frequency signal being allowed to pass through.In one embodiment,
Low pass mean filter 110 executes Fuzzy function to brightness component 104.
For the image of intermediate complexity, the filtering if necessary to moderate and/or the system resource if there is average magnitude
It can use, then median filter 112 can be used to filter brightness component 104.As the skilled person will appreciate,
Median filter 112 handles brightness component 104 pixel by pixel, and each pixel is replaced with the pixel and its surrounding pixel
Intermediate value.For example, around pixel of interest 3 × 3 pixel window (i.e. 9 pixels in total) can be considered in median filter 112.
Median filter 112 sorts 9 pixels by its brightness-value, selects the value in intermediate (that is, the 5th) position, and will be closed
The pixel of note replaces with selected value.In one embodiment, filter 112 is the median filter that sorts or sort, and can
With the pixel in position (for example, the 3rd or the 6th position) any in the pixel list of selected and sorted.In one embodiment, such as
Absolute difference selected by fruit between value and initial value is greater than threshold value, then retains initial value；If this difference is less than or equal to threshold value, assign
The value of sequence.
Then may be used for the image of high complexity if necessary to a large amount of filtering and/or if there is a large amount of system resources are available
To use sef-adapting filter 114 to filter brightness component 104.Sef-adapting filter 114 is bright based on what is be dynamically determined
Spend the feature of component 104 to select filtering technique, it is as follows to make an explanation in more detail.
Low pass mean filter 116 (for example, 5 × 5 low pass mean filters) can be used to filter color component 106
Wave.In one embodiment, color component 106 is complicated not as good as brightness component, and/or influenced by low optical noise it is smaller, and
Thus less filtering is needed.Filter 116 can be with absolute difference summation (SAD) time average filter or it is any its
The similar filter of his type.System 100 may include more than one color component filter 116, and multiple colors are divided
Measuring one of filter 116 can complexity based on color component 106, the availability of system resource and/or desired
Filter quality level selects.
Fig. 2 diagram is for removing the flow chart 200 of noise from low light image.First is applied to the luminance component of low light image
Filter (step 202), and to the chromatic component application second filter (step 204) of low light image.It will be filtered
Luminance component is combined with filtered chromatic component, to generate filtered low light image (step 206).First filter
It can be low pass mean filter 110, intermediate value/sequence median filter 112, or based on the adaptive filter of edge/Gaussian Profile
Wave device 114, as described above, and second filter can be low pass or time average filter 116.
Fig. 3 is one of sef-adapting filter 114 and realizes 300 diagram.302 detection image 102 of edge difference filter
Luminance component 104 in edge.Edge difference filter 302 is referred to as difference filter.Edge difference filter 302
It can detecte the edge in luminance component 104, while retaining high frequency detail therein.Edge detection process will be in luminance component
Pixel is divided into edge pixel and non-edge pixels.
Based on the filter 304 of expansion by the way that the distribution of results of edge detection to adjacent pixel to be modified to the filter of edge difference
The output of wave device 302.The filter based on expansion can be modified to be convenient to realize in for example embedded and/or DSP platform.
For example, if 4 pixels in expansion a line, can according to the position of pixel by this 4 pixel shifts, with word boundary pair
Together.In various embodiments, the filter 304 based on expansion is morphological filter, 3 × 4 expansion filters or 4 × 3 expansion filters
Wave device.The region for being designated as the pixel of edge pixel can be expanded or be expanded to be incorporated to it by the filter 304 based on expansion
His adjacent pixel.For example, the pixel with the different shading value of pixel adjacent thereto may be caused by low optical noise, but if
The edge for being located proximate to detect of the pixel, then the pixel may be the knot of the real physical characteristics of institute's captured image instead
Fruit.Such pixel for being occurred based on the filter 304 of expansion by the edge that will be close to detect is related to edge pixel to prevent
It is only mistakenly appointed as to the pixel of noise generation.
Then, in the luminance component 104 of control adjacent pixel regions (for example, 3 × 3 adjacent block of pixels) analysis expansion
Each non-edge pixels.Depending on such as passing through between the pixel the analyzed pixel adjacent thereto that Gaussian Profile engine 306 calculates
Difference, assign new value to the pixel according to assignment unit 308-312, and exported by output unit 314.
More specifically, Gaussian Profile engine 306 calculates the Gaussian Profile of block or window around analyzed pixel
Average value and variance.The deviation of the average value of the pixel and block is calculated, and by it compared with variance.If pixel and variance it
Between difference be much larger than average value (for example, bigger than the three times of standard deviation), then the pixel may be caused by low optical noise.In this feelings
In condition, which is replaced with the intermediate value of block of pixels by intermediate value block 308.If the nearly average value of differential between pixel and variance,
The pixel analyzed is replaced with the result that low-pass filtering is carried out to block of pixels by low-pass filter 310.If pixel and variance it
Between difference be less than average value, then the pixel analyzed is transmitted to output block 314 with remaining unchanged by block of pixels 213.
Generally speaking, the algorithm that assignment unit 308-312 is used can be summarized by following equation:
{ if (pixel of analysis)-(average value of block of pixels) } > N × (variance of block of pixels):
Output=block of pixels intermediate value (1)
{ if (pixel of analysis)-(average value of block of pixels) } > M × (variance of block of pixels):
The result (2) of output=block of pixels low-pass filtering
{ if (pixel of analysis)-(average value of block of pixels) } > P × (variance of block of pixels):
Output=original analysis pixel (3)
Wherein P≤M≤N.That is, assigning intermediate value 308 to output 314 for big difference, for medium difference, low pass is assigned
Filter 310 as a result, and for small difference, retaining original pixel 312.In one embodiment, pass through the hard of particular allocation
Part executes the operation that above equation (1)-(3) execute.In another embodiment, median operation is executed by median filter 112,
And low-pass filtering is executed by low pass mean filter 110, as shown in Figure 1.
Fig. 4 depicted exemplary luminance component 400.Edge 402 is detected between image-region 404 and 406.As described above
, edge pixel can be appointed as by the pixel 408 that the filter 304 based on expansion will be close to edge 402.It can analyze first
Pixel 410, and by it compared with its 3 × 3 surrounding pixel 412.In this case, because of the pixel 410 and block of pixels of analysis
Difference between 412 average value is much larger than the variance of (that is, being greater than threshold value N) block of pixels 412 (that is, the pixel adjacent thereto of pixel 410
There are big differences between 412 brightness value), then pixel 410 is replaced with to the intermediate value of 3 × 3 surrounding pixel 412.
In another example, one other pixel 414 is analyzed, and by it compared with its surrounding pixel 416.Herein, because
When compared with the variance of block of pixels 412, the difference between the pixel 414 of analysis and the average value of block of pixels 412 is less than the first threshold
Value N but it is greater than second threshold M, so pixel 414 is replaced with the result to 416 low-pass filtering of block.Finally, because with picture
When the variance of plain block 420 compares, the difference between the pixel 418 of third analysis and the average value of its surrounding pixel block 420 is much smaller than
Threshold value P, so pixel 418 remains unchanged.
In one embodiment, system as described above 300 analyzes each pixel in luminance component 104.In other realities
It applies in example, system 300 only analyzes a subset of the overall pixel in luminance component 104.For example, system 300 can be analyzed only
The pixel (for example, every pixel of a pixel) of even-numbered in luminance component 104.It can will analyze the picture of even-numbered
The result of element is not only applied to the pixel itself, but also be applied to adjacent odd number pixel (for example, in a line with institute
The adjacent pixel of the pixel of the even-numbered of analysis).Because two pixels are adjacent, the result calculated a pixel may
It is similar to the uncalculated result of adjacent pixel, and the result of the pixel of analysis may be generated only applied to the two pixels
Small error.It can choose other pixel subset (such as odd pixel, the pixel every N-1 pixel, diagonal pixels or multiple
The pixel of row/column) for analyzing.In example above, the pixel of analysis may be constructed the 50% of total pixel, or may be constructed total
Any other percentage of pixel.
Fig. 5 is the flow chart 500 of method of the diagram for adaptively filtering out noise from low light image.Using for example above
Edge swell (the step that the edge difference filter 302 of description and filter 304 based on expansion will detect in image
502).Edge detection and expansion divide an image into edge pixel and non-edge pixels, and by non-edge pixel with
Compare (step 504) in region around pixel.According to the result of the comparison, described above, non-edge pixels is optionally replaced
(step 506).
Fig. 6 is the block diagram 600 for removing the system of noise from low light image and dividing an image into subregion.It draws
Parallel circuit 602 divides an image into two or more regions, and luminance component application of the filter circuit 604 to each region
First filter.Once filtering each region, then circuit 606 is recombinated by each filtered region combination to create warp
Cross the image of filtering.Generally speaking, these regions can be the size of any M × N, such as 16 × 16 pixels.
In one embodiment, system 600 can be used for dividing an image into the quantity pair with available filters circuit 604
The region for the quantity answered.Each filter circuit 604 may include the system 100 for removing low optical noise from each region,
As shown in Figure 1.Filter circuit 604 may include first filter for filtering to luminance component and for chromatic component
The second filter of filtering.It is then possible to concurrently be filtered simultaneously to multiple regions, to reduce to needed for whole image filtering
Time.In other embodiments, the quantity in region is greater than the quantity of filter circuit 604, and concurrently handles some areas
Domain, and other regions are lined up.
In another embodiment, each image-region is sequentially handled using only a filter circuit 604.It is real herein
It applies in example, image can be defined by the ability of the amount and/or filter circuit 604 of available memory or other memory spaces
The size in region.The size in region can be adjusted to consume more or fewer resources, this depends on the constraint of specific application.
For example, the application with very limited memory may need small region.It can store the row with management region or image
It is mobile to be easy to data in switching and/or combined image area with the historical information of column.
Fig. 7 diagram is for by dividing an image into subregion to remove the method 700 of noise from low light image.By image
It is divided into multiple regions (step 702), and to the luminance component in each region (sequentially or concurrently) using first filter (step
It is rapid 704).The region filtered respectively is combined into filtered image (step 706).
It may include low-pass filtering being carried out to region, to region progress median filtering and/or to area using first filter
Domain is adaptively filtered, such as described in reference diagram 1 above.Sef-adapting filter by region pixel and adjacent pixel ratio
Compared with, and optionally replace it.It is for another example described above, the chromatic component of image can also be resolved into figure by division circuit 602
As region, it is filtered with second filter, and is recombinated by recombination circuit 606.The image-region of chromatic component
Size can be identical or different with the size of the image-region of luminance component.In one embodiment, image procossing as a whole
Chromatic component because it is with lower complexity, and luminance component is divided and is respectively handled.
The embodiment of the present invention can be used as hardware, software and/or firmware to provide.For example, system 100,300,600 can
To realize in embedded equipment (such as ASIC, FPGA, microcontroller or other similar device), and video can be included in
Or in camera.In other embodiments, the element of system 100,300,600 can be realized in software, and be included in
On desktop PC, notebook computer, netbook computer or handheld computer.In these embodiments, online camera shooting
Head, cellular telephone camera or other similar device can capture image or video, and system 100,300,600 can be from it
It is middle to remove low optical noise.The present invention is also used as on one or more manufactures or one or more computers of middle embodiment
Readable program provides.The manufacture can be any suitable hardware device, such as floppy disk, hard disk, CD optical ROM, DVD
Optical ROM, Blu-ray Disc, flash card, PROM, RAM, ROM or tape.Generally speaking, computer-readable program can use and appoint
What programming language is realized.The some examples for the language that can be used include C, C++ or JAVA.It can also be by software program into one
Step is construed to machine language or virtual machine instructions, and stores it in program file in this manner.It is then possible to by program text
Part be stored on one or more manufactures or in which.
Described above is certain embodiments of the present invention.But it will be clearly note that the present invention is not limited to these implementations
Example, but the addition and modification to content explicitly described herein are also intended to and are comprised in the scope of the present invention.Moreover, to manage
The feature of various embodiments described herein and non-exclusive is solved, and can be existed in the form of multiple combinations and arrangement,
Even if such combination or arrangement do not provide clearly herein, without departing from the spirit and scope of the invention.In fact, not carrying on the back
Under the premise of the spirit and scope of the present invention, those of ordinary skill in the art by be susceptible to content described herein variation,
Modification and other realizations.Therefore, the present invention should not only be described by being described above property to define.
Claims (25)
1. a kind of for filtering out the sef-adapting filter of noise from low light image, the sef-adapting filter includes:
Morphological filter is used for the edge swell that will be detected in described image, to divide the image into non-side
The fringe region in edge region and expansion；And
Compare filter, is used for by the pixel in the non-edge compared with the region around the pixel, Yi Jiji
In comparison result, it is determined whether replacing with the pixel and is at least partly obtained from the region around the pixel
New pixel.
2. sef-adapting filter according to claim 1 further includes difference filter, the difference filter is for examining
Survey the edge in described image.
3. sef-adapting filter according to claim 1, wherein the morphological filter includes expansion filter.
4. sef-adapting filter according to claim 3, wherein the expansion filter is 3 × 4 expansion filters.
5. sef-adapting filter according to claim 1, wherein the region around the pixel corresponds to 3 × 3 pictures
Plain region.
6. sef-adapting filter according to claim 1, wherein the sef-adapting filter is by the pixel and the picture
Difference between the average value in the region around plain is compared with the variance in the region around the pixel.
7. sef-adapting filter according to claim 6, wherein the variance is the region around the pixel
The variance of Gaussian Profile.
8. sef-adapting filter according to claim 7 further includes for calculating the region around the pixel
The Gaussian Profile the variance circuit.
9. sef-adapting filter according to claim 6, wherein the average value is the region around the pixel
Gaussian Profile average value.
10. sef-adapting filter according to claim 7 further includes for calculating the region around the pixel
The Gaussian Profile the average value circuit.
11. sef-adapting filter according to claim 1, wherein the relatively filter is based at least partially on the picture
The intermediate value in the region around plain obtains the new pixel.
12. sef-adapting filter according to claim 1, wherein the relatively filter is based at least partially on the picture
The low-pass filter values in the region around plain obtain the new pixel.
13. sef-adapting filter according to claim 1, wherein the relatively filter is by adjacent with the pixel the
Two pixels replace with the new pixel.
14. sef-adapting filter according to claim 13, wherein the pixel is even pixel, and second picture
Element is odd pixel.
15. sef-adapting filter according to claim 1, wherein the relatively filter retains the pixel.
16. a kind of method for adaptively filtering out noise from low light image, which comprises
The edge swell that will be detected in the picture, to divide the image into the marginal zone of non-edge and expansion
Domain；
By the pixel in the non-edge compared with the region around the pixel；And
Based on comparative result, it is determined whether replace with the pixel at least partly from the region around the pixel
Obtained new pixel.
17. according to the method for claim 16, further including the edge detected in described image.
18. according to the method for claim 16, wherein the pixel is wrapped compared with the region around the pixel
It includes the region around the difference and the pixel between the variance in the region around the pixel and the pixel
Average value compares.
19. according to the method for claim 18, wherein the variance is the Gauss point in the region around the pixel
The variance of cloth.
20. according to the method for claim 18, wherein the average value is the Gauss in the region around the pixel
The average value of distribution.
21. according to the method for claim 16, wherein the new pixel is at least partially based on around the pixel
What the intermediate value in the region obtained.
22. according to the method for claim 16, wherein the new pixel is at least partially based on around the pixel
What the low-pass filter values in the region obtained.
23. according to the method for claim 16, further including that the second pixel is replaced with to the new pixel, wherein described the
Two pixels are adjacent with the pixel.
24. according to the method for claim 23, wherein the pixel is even pixel, and second pixel is odd number
Pixel.
25. according to the method for claim 16, wherein the relatively filter retains the pixel.
Applications Claiming Priority (7)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US12/950,666 | 2010-11-19 | ||
| US12/950,671 | 2010-11-19 | ||
| US12/950,671 US20120128244A1 (en) | 2010-11-19 | 2010-11-19 | Divide-and-conquer filter for low-light noise reduction |
| US12/950,664 US8755625B2 (en) | 2010-11-19 | 2010-11-19 | Component filtering for low-light noise reduction |
| US12/950,664 | 2010-11-19 | ||
| US12/950,666 US8699813B2 (en) | 2010-11-19 | 2010-11-19 | Adaptive filter for low-light noise reduction |
| CN201180062954.XA CN103270746B (en) | 2010-11-19 | 2011-11-15 | Component for low smooth noise reduction filters |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201180062954.XA Division CN103270746B (en) | 2010-11-19 | 2011-11-15 | Component for low smooth noise reduction filters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105657216A CN105657216A (en) | 2016-06-08 |
| CN105657216B true CN105657216B (en) | 2019-07-05 |
Family
ID=45375477
Family Applications (3)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610649854.1A Active CN106161878B (en) | 2010-11-19 | 2011-11-15 | system, method and non-volatile medium for removing noise from an image |
| CN201610124067.5A Active CN105657216B (en) | 2010-11-19 | 2011-11-15 | Component for low smooth noise reduction filters |
| CN201180062954.XA Active CN103270746B (en) | 2010-11-19 | 2011-11-15 | Component for low smooth noise reduction filters |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610649854.1A Active CN106161878B (en) | 2010-11-19 | 2011-11-15 | system, method and non-volatile medium for removing noise from an image |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201180062954.XA Active CN103270746B (en) | 2010-11-19 | 2011-11-15 | Component for low smooth noise reduction filters |
Country Status (4)
| Country | Link |
|---|---|
| EP (1) | EP2641390B1 (en) |
| KR (1) | KR101537295B1 (en) |
| CN (3) | CN106161878B (en) |
| WO (1) | WO2012068085A2 (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8755625B2 (en) | 2010-11-19 | 2014-06-17 | Analog Devices, Inc. | Component filtering for low-light noise reduction |
| GB201314068D0 (en) * | 2013-08-06 | 2013-09-18 | Microsoft Corp | Encoding Video Captured in low light |
| US9990730B2 (en) | 2014-03-21 | 2018-06-05 | Fluke Corporation | Visible light image with edge marking for enhancing IR imagery |
| US10152811B2 (en) | 2015-08-27 | 2018-12-11 | Fluke Corporation | Edge enhancement for thermal-visible combined images and cameras |
| KR102383578B1 (en) * | 2015-09-09 | 2022-04-08 | 삼성전자주식회사 | An imaging processing apparatus, a decoding apparatus and a method for processing an image |
| KR101802035B1 (en) | 2015-11-03 | 2017-12-20 | 한국과학기술원 | Ct image denoising framework considering the noise characteristics |
| DE102015223175A1 (en) * | 2015-11-24 | 2017-05-24 | Conti Temic Microelectronic Gmbh | Driver assistance system with adaptive environment image data processing |
| KR101961816B1 (en) * | 2017-11-30 | 2019-03-25 | 전남대학교산학협력단 | Directional average filtering method that can denoise and edge preservation of image |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6229578B1 (en) * | 1997-12-08 | 2001-05-08 | Intel Corporation | Edge-detection based noise removal algorithm |
| CN101087365A (en) * | 2006-06-10 | 2007-12-12 | 中兴通讯股份有限公司 | A method for filtering image mixed noise |
Family Cites Families (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004236251A (en) * | 2003-02-03 | 2004-08-19 | Noritsu Koki Co Ltd | Image processing apparatus, image processing method, image processing program, and recording medium with the image processing program recorded thereon |
| US7724307B2 (en) * | 2004-07-28 | 2010-05-25 | Broadcom Corporation | Method and system for noise reduction in digital video |
| CN100562129C (en) * | 2004-11-05 | 2009-11-18 | 中兴通讯股份有限公司 | The time-domain filtering method of video coding |
| CN100420279C (en) * | 2005-03-31 | 2008-09-17 | 北京中星微电子有限公司 | Method and apparatus for lowering image noise under low illumination |
| KR100776389B1 (en) * | 2005-12-29 | 2007-11-16 | 엠텍비젼 주식회사 | Apparatus for color interpolation by using adjustable threshold |
| EP1969559B1 (en) * | 2005-12-30 | 2018-09-05 | Telecom Italia S.p.A. | Contour finding in segmentation of video sequences |
| KR100780242B1 (en) * | 2006-11-14 | 2007-11-27 | 삼성전기주식회사 | Method and apparatus for removing noise in dark area of image |
| KR100872253B1 (en) * | 2007-08-23 | 2008-12-05 | 삼성전기주식회사 | Method for eliminating noise of image generated by image sensor |
| CN101197934B (en) * | 2007-12-21 | 2010-06-23 | 北京中星微电子有限公司 | Method and device for reducing noise between frames |
| KR101442153B1 (en) * | 2008-01-15 | 2014-09-25 | 삼성전자 주식회사 | Method and system for processing for low light level image. |
| CN101237523B (en) * | 2008-02-25 | 2013-03-06 | 宝利微电子系统控股公司 | Main edge detection method and noise reduction method and device based on this method |
| US8149336B2 (en) * | 2008-05-07 | 2012-04-03 | Honeywell International Inc. | Method for digital noise reduction in low light video |
| US8866936B2 (en) * | 2008-07-24 | 2014-10-21 | Florida State University of Research Foundation | Systems and methods for training an active random field for real-time image denoising |
-
2011
- 2011-11-15 KR KR1020137015695A patent/KR101537295B1/en active IP Right Grant
- 2011-11-15 CN CN201610649854.1A patent/CN106161878B/en active Active
- 2011-11-15 WO PCT/US2011/060756 patent/WO2012068085A2/en active Application Filing
- 2011-11-15 EP EP11799339.4A patent/EP2641390B1/en active Active
- 2011-11-15 CN CN201610124067.5A patent/CN105657216B/en active Active
- 2011-11-15 CN CN201180062954.XA patent/CN103270746B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6229578B1 (en) * | 1997-12-08 | 2001-05-08 | Intel Corporation | Edge-detection based noise removal algorithm |
| CN101087365A (en) * | 2006-06-10 | 2007-12-12 | 中兴通讯股份有限公司 | A method for filtering image mixed noise |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103270746B (en) | 2016-09-07 |
| CN105657216A (en) | 2016-06-08 |
| WO2012068085A3 (en) | 2012-08-09 |
| EP2641390B1 (en) | 2017-11-08 |
| CN103270746A (en) | 2013-08-28 |
| EP2641390A2 (en) | 2013-09-25 |
| KR101537295B1 (en) | 2015-07-16 |
| KR20130136486A (en) | 2013-12-12 |
| CN106161878B (en) | 2019-12-10 |
| WO2012068085A2 (en) | 2012-05-24 |
| CN106161878A (en) | 2016-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105657216B (en) | Component for low smooth noise reduction filters | |
| Kim et al. | Deep learning of human visual sensitivity in image quality assessment framework | |
| US8755625B2 (en) | Component filtering for low-light noise reduction | |
| Malinski et al. | Fast adaptive switching technique of impulsive noise removal in color images | |
| US8699813B2 (en) | Adaptive filter for low-light noise reduction | |
| CN103927734B (en) | A kind of based on the broad image quality evaluating method without reference | |
| US20120128244A1 (en) | Divide-and-conquer filter for low-light noise reduction | |
| Sandić-Stanković et al. | Fast blind quality assessment of DIBR-synthesized video based on high-high wavelet subband | |
| Hasikin et al. | Adaptive fuzzy intensity measure enhancement technique for non-uniform illumination and low-contrast images | |
| Chetouani | Image quality assessment without reference by mixing deep learning-based features | |
| Rao et al. | Illumination-based nighttime video contrast enhancement using genetic algorithm | |
| Agrawal et al. | A joint cumulative distribution function and gradient fusion based method for dehazing of long shot hazy images | |
| Ponomaryov et al. | Fuzzy color video filtering technique for sequences corrupted by additive Gaussian noise | |
| JP5097141B2 (en) | Noise reduction device and noise reduction program | |
| JP2007048104A (en) | Image processing method, image processor and image processing program | |
| JP2008294706A (en) | Video signal processing device, video signal processing method, program of video signal processing method, and recording medium with recorded program of video signal processing method | |
| CN111147693B (en) | Noise reduction method and device for full-size photographed image | |
| Almaghthawi et al. | Differential evolution-based approach for tone-mapping of high dynamic range images | |
| Neelima et al. | Performance Evaluation of Clustering Based Tone Mapping Operators with State-of-Art Methods | |
| Ghous et al. | Efficient image enhancement using improved RIQMC based ROHIM model | |
| Dubey et al. | A survey paper on noise estimation and removal through principal component analysis | |
| Jamali et al. | Weighted fuzzy-based psnr for watermark visual quality evaluation | |
| US20100054597A1 (en) | Image processing and generation of focus information | |
| Song et al. | No-reference quality assessment of night-time images via the analysis of local and global features | |
| JP7496935B2 (en) | IMAGE PROCESSING APPARATUS, IMAGE PROCESSING METHOD, AND IMAGE PROCESSING PROGRAM |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |